Brit. J. industr. Med., 1955, 12, 103.

*DIURNAL VARIATION IN MENTAL PERFORMANCEA STUDY OF THREE-SHIFT WORKERS

BY

BO BJERNER, AKE HOLM, and AKE SWENSSONFrom Division of Occupational Medicine, Aarolinska Hospital, Stockholm, Sweden

(RECEIVED FOR PUBLICATION MARCH 30, 1954)

A great deal of research has been done on thediurnal variations in different forms of animal andhuman activity. The first diurnal variation to beobserved was the one in body temperature, and itis the one most widely recognized today. Similardiurnal variations have been observed in a series ofother physical processes. Extensive surveys of theliterature have been given by Holmgren (1936),Jores (1937), and Kleitman (1939, 1949).A series of investigations has also been carried

out on diurnal variations in mental performance.At first these investigations were done by psycho-logists independently of physiological research.Freeman and Hovland (1934) gave an exhaustivereview of the literature and assembled the contri-butions of different workers in a table in which theforms of variation observed were divided intodifferent classes. As Kleitman (1939) remarked," No matter what curve one obtains, there is aconvenient niche in Freeman and Hovland's tablefor it ". The main reason for the variety of curvesis that most of the series studied are so small thatchance variations have played too great a part.Kleitman and his associates made a thorough studyof the variations in mental performance duringwaking hours. Kleitman stated that all the curvesthey obtained fitted into class C of Freeman andHovland, namely, " a peak in the middle of thewaking period ". He believed that the fluctuationsin mental performance were caused by variationsin body temperature. In 1949 one of us studied thereaction time in five subjects kept awake for 24hours and longer (Bjerner, 1949). Two of themshowed a regular curve with a peak in the length ofreaction times and a number of false reactions at4 a.m. The other three showed longer reactiontimes during the night but no distinct diurnalvariation. This was the first study of this kind tobe made during the night.

* The term " diurnal " is used here in the sense of Kleitman torefer to the 24-hour cycle of day and night.

Another branch of these investigations, apparentlydone independently of the two others, is the researchbegun about 1900 on the spontaneous activity ofanimals during day and night. A distinct diurnalvariation has been observed, which, however, iseasily influenced by external stimuli such as lightand darkness. If the night is made light and theday dark, the animal quickly inverts its rhythm ofactivity. This takes place by a gradual displacementof the activity, in rats completed within a weekafter the change in illumination (Holmgren andSwensson, 1953).

In general it may be said that studies of thediurnal variation of different functions have some-times given greatly conflicting results and thetheoretical aspects of the problem have not as yetbeen satisfactorily elucidated. Obviously, however,variations of this kind must be of great significancein everyday life, especially when it comes to nightwork and/or shift work. But, very few studies havebeen done on this particular aspect of the problemand none of them has given results of any greatpractical importance.The most extensive studies were done in England

during and after the first world war. The resultsare given in the Final Report of the Health ofMunition Workers Committee (Ministry of Muni-tions, 1918) and in reports of the IndustrialFatigue Research Board (1919-22). They have beensummarized by Vernon (1921, 1940).One of the main objects of the Health of Munition

Workers Committee was to study whether therewas any difference in working efficiency during dayand night. Several different series of investigationswere done and it was found that sometimes theproduction was higher in night shifts and sometimeshigher in day shifts.

It was observed that when women had mono-tonous work, requiring little physical effort, theproduction in discontinuous night work aboutequalled that for the same kind of work during the

103

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

BRITISH JOURNAL OF INDUSTRIAL MEDICINE

day, or lay at the most 10% under. When it cameto continuous night work, on the other hand, thenightly production fell distinctly below that forday-time work. The results for men were about thesame. The Health of Munition Workers Committeeused as a criterion the total production per shift.The Industrial Fatigue Research Board tried to

get an idea of the variation in production from hourto hour during day and night shifts. The mostextensive investigations in this connexion were doneby Vernon with workers in steel mills. But theseinvestigations did not reveal any difference betweenday and night work, either as regards work requiringa large amount of physical effort or light super-visory work.Vernon also studied the accident rate at

different hours of the day and night in two largeseries. His figures showed a lower frequency ofaccidents during the night than during the day.Discussing his results as regards accident rate andproduction at different hours, Vernon came to theconclusion that both were dependent upon manyother factors than fatigue and therefore could notbe taken as indicators of the degree of fatigue.

After the discouraging results of these largeinvestigations, nothing more was done in this field.It is only recently that the problem has been takenup again. Other angles of approach are now beingtried.Browne (1949), studying telephone operators,

calculated the influence of the hour on the length oftime taken to answer incoming calls. He foundthat, on the average, the longest reaction timeoccurred between 3 and 4 a.m. This was true evenwhen there were more than 80 calls per hour,showing that the long reaction time was not due tothe operator dozing between calls.

Menzel (1950) examined the frequency of accidentsfor which a penalty was inflicted in a group ofrailway employees working in three shifts beginningat 6 a.m., 2 p.m., and 10 p.m. He found that thefrequency was greatest between 10 p.m. and 2 a.m.When it is desired to study diurnal variations in

mental performance in industrial work, greatdifficulties are encountered. The greatest of theselies in the fact that the working conditions aregenerally not the same during the night as duringthe day. Efficiency is also influenced by a series ofother factors, as, for example, setting of a piecerate, which could easily conceal a diurnal variationin mental performance. Large series are needed inorder to eliminate chance variation.The investigations of night and shift work

described in the present article were done in 1946.We tried to avoid the sources of error just spoken

of when choosing study material. Preliminaryreports were presented in 1948, 1949, and 1953(Bjerner, ¶{olm, and Swensson, 1948; Bjerner, 1949;Bjerner and Swensson, 1953).

Present SeriesThe bulk of the material comes from a large

gasworks (I) in southern Sweden. The subjectswere men whose work it was to enter figures incolumns in ledgers. Some of the figures were takendirectly from meters and some were the results ofcomputation. The first type was taken from theinstruments which registered the consumption ofgas, amount of gas in the gas tanks, gas pressure,temperature, height of the barometer, and otherconditions. The figures calculated were those forthe amount of gas produced and consumed. Theywere obtained by additions and subtractions ofnumbers, generally three or four digits in length.They were entered every hour during the day andnight, and it took an experienced person 10 to 15minutes to put them all down.

Every morning at six o'clock the figures weregone through and any errors in calculation ornotation were corrected. We are convinced afterour visit to the gasworks that all the figures werestrictly checked.The men worked in shifts of eight hours each,

beginning at 6 a.m., 2 p.m., and 10 p.m. Shiftswere changed every week. The working team wasusually made up of three men. After 1920 the mengot a free day each week, and a substitute wasobtained for these days.We examined the ledgers filled in between 1901

and 1943. We made an analysis of the errors in thefigures, both the errors in noting the readings fromthe instruments and in calculation. The figures wereentered in pencil. Mistakes were corrected in oneof two ways. Sometimes the wrong number wasrubbed out and the right one substituted. Othertimes the wrong number was crossed out and theright one written above it. The paper had a shinysurface and when it was held so that the light fellobliquely on it, it was easy to detect the erasures.We paid no attention to the manner of correction.Mistakes in two or more digits of one number werecounted as one error only. Errors in differentcolumns were counted separately, even when madein the same entry.The ledgers were gone through and all the errors

ticked off. Then the different errors were trans-ferred to different tables of distribution. The tableswere compiled in a way that made it possible tocalculate how the errors were distributed throughoutthe day and night, week, month, and year. In

104

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

DIURNAL VARIATION IN MENTAL PERFORMANCE 105

TABLE 1

FREQUENCY OF ALL ENTRIES WITH ERRORS DISTRIBUTED ACCORDING TO TIME OF DAY AND NIGHT

- Observations with Errors TotalHour (Number of Errors) No.

io 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Errors

06 2,134 339 60 11 2 - 3 - 2 1 1 1 - _ - 56407 2,173 318 49 9 2 1 1 - - 1 - - _ - - 47108 2,209 290 47 7 - - - - - - -_ - 41009 2,184 318 36 6 4 3 1 2 - - - - - _ - 45910 2,1491 334 55 9 6 - - 1 - - - - _ - 50211 2,172 298 61 16 3 3 - 1 - - - - _ - - 50212 2,098 364 68 21 2 - 1 _ - _ _ 57713 2,177 313 41 15 1 3 2 1 _ 1 - - _ _ - 48714 2,100 373 64 12 2 - - 1 2 - - - - _ - 56815 2,127 333 73 17 2 - - 1 1 - - - - - - 55316 2,145 339 57 7 4 1 - - 1 - - - _ - 50317 2,137 348 53 13 1 - 1 - - - 1 - - - - 51318 2,130 355 55 8 2 2 - 1 - - 1 - - - 52519 2,165 318 56 11 2 1 - - 1 - _ _ _ _ - 48420 2,186 296 52 14 3 1 1 - - - 1 - _ _ - 47521 2,120 353 58 12 5 1 - 2 - - - 2 1 - - 57822 2,121 350 65 13 1 1 - - 2 - - 1 - - - 55523 2,099 362 58 24 4 2 2 2 - - - - - 1 - 61524 2,083 367 74 17 7 4 - 1 1 - - - - - - 62901 1,997 421 89 31 10 2 1 1 1 - _ 1 - - - 77402 1,931 483 97 27 9 4 2 - 1 - - - - 83403 1,795 563 147 29 12 2 - - 5 _ - - - 1 1,05604 1,878 517 110 32 9 3 1 - 1 I 1 90105 2,006 416 108 17 3 2 1 - - - -- 725

Totals 50,316 8,768 1,633 378 96 37 17 14 18 4 4 6 1 2 2 14,289

Total observations: 61,296; Mean number of errors per observation = 0 2331 (S.D.=0-36880)

addition, separate calculations were made for theerrors made in copying down from the instrumentsand the errors in figures based on calculation.

It was possible to distinguish the handwriting ofseveral different workers, and so to study thedistribution of errors for separate men. The resultsof three of the workers, called A, C, and D, wereanalysed in detail.

For the final investigation the years 1912 to 1931were chosen. The ledgers for these years were inexcellent condition and no pages had to be dis-carded. Apart from the work already done for thepreliminary investigation, the ticking off andtabulation of this material was all done by oneperson. This person was not acquainted with theresults of the preliminary investigation.

In order to check the results, similar investigationson a smaller scale were carried out at anothergasworks (II) and at a paper mill.

Results of Preliminary InvestigationThe ledgers for theyears studied showed that 14,289

errors were committed in 61,296 entries. Table 1shows the distribution of the single and multipleerrors in the 24 hours of the day and night. Asseen there, there were many entries containingseveral mistakes. Analysis revealed that many ofthe multiple errors occurred in the columns forthe pressure in the different gas-tanks. Study atthe gasworks showed that in certain situations all

the pressure meters changed figures at the sametime. If this happened when a man was reading offthe figures, he had to rub out the whole series offigures he had just written. Because of this, thesecolumns were excluded as errors from the analysis.The results when this was done are seen in Table 2.

TABLE 2

FREQUENCY OF ENTRIES (EXCLUDING THOSE WITHPRESSURE METER CHANGES) WITH ERRORS DISTRI-

BUTED ACCORDING TO TIME OF DAY AND NIGHT

Observations with Errors TotalHour No. of

0 1 2 3 4 5 6 Errors

06 2,297 236 19 1 1 - - 28107 2,359 178 13 3 1 - - 21708 2,358 184 10 2 - - - 21009 2,373 169 11 1 - - - 19410 2,364 174 12 4 - - - 21011 2,381 156 14 2 1 - - 19412 2,357 181 14 1 1 - - 21613 2,366 172 16 - - - - 20414 2,315 208 28 - 2 - 1 27815 2,348 185 16 3 1 1 - 23516 2,364 173 15 2 - - - 20917 2,345 189 16 4 - - - 23318 2,351 182 21 _- - - 22419 2,362 175 12 5 - - - 21420 2,394 146 11 3 - - - 17721 2,372 162 19 1 - - - 20322 2,325 197 28 4 - - - 26523 2,332 193 25 3 1 - - 25624 2,349 181 19 4 1 - - 23501 2,256 251 39 7 1 - - 35402 2,240 267 38 9 - - - 37003 2,135 352 62 5 - - - 49104 2,158 341 50 4 1 - - 45705 2,240 276 33 4 1 - - 358

Totals 55,741 4,928 541 72 12 1 1 6,285

Mean no. of errors Per observation = 0-1025 (S.D. = 0-1199)

c

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

BRITISH JOURNAL OF INDUSTRIAL MEDICINE

TABLE 3

DISTRIBUTIONS ON DAYS OF THE WEEK OF ENTRIES WITH ERRORS

No. of Errors Frequency Frequency Calculated According to: Frequency Frequency Calculated According to:No.ofEfforsObserved Poisson Greenwood-Yule Observed Poisson I Greenwood-Yule

Mondays Tuesdays0 7931 7847 5 7933-2 7961 7926-4 7963-31 731 863-2 725 9 702 792*6 698-72 83 47 5 87-7 86 39*6 85-03 13 1-7 11-4 10 1*3 11*34 2 0-1 1 5 0 0.1 1o65 ---0 - 0-26 _ _ _ 1 _ _

Total 8760 - _ 8760 --n 1 1 - 1 1X2 _ 50-18 0-61 _ 86-87 0-37P _ <0-001 0 43 - <0-001 0-54PN - - 0-89 _ - 0.95

Wednesdays Thursdays0 8033 7948-1 8035 1 7858 7826-0 7857-41 679 794-8 669-5 785 860-9 788-62 58 48 5 69-9 84 47-4 80*83 14 1-3 9-5 7 1-7 8-34 _- - 1 005 09

Total 8784 _ _ 8736 - -

n 1 1 - 1 1X2 _ 27 53 4-29 _ 44-31 0-32P _ <0-001 0 04 - <0-001 0-57PN - - 0-24 _ - 0.95

Fridays Saturdays0 7982 7962-2 7981*41 7979 7869-0 7984-61 659 716-5 660-89 703 865-6 687-02 65 32-2 62-80 81 47-6 94-83 5 1-0 6-22 16 1-8 14-74 1 - 0-63 5 - 2-4

Total 8712 - - 8784 -

n 1 1 - 1 1x2 _ 47-34 0-20 _ 88-11 3-03P - <0-001 0-65 _ <0-001 0-08PN - _ 0-98 _ _ 0 40

671

83

7

7926-38792-6439-631-320-030-0008

7993-76672-2581-4110-86

1 520-22

Total 8760 _n - 1 2X2 - 70.97 0-23P - <0-001 0-63PN - - 0-97

1

The errors are not distributed according to Poissonbut the distribution tallies fairly well with the curve

constructed by Greenwood and Yule (1920) forthe distribution of accidents among differentpersons. As seen from Table 3, the distribution ofcertain sections of this series fits the Greenwood-Yule curve remarkably well for each day of the week-much better than the distributions noted in thestudies of accidents by Greenwood and Yule andby Newbold (1926).

Because of this great stability in the series we

felt justified in studying the total number of errors

at different times instead of the frequency of hourswith and without errors.The last column of Table 2 shows a distinct

diurnal variation with a peak at 3 a.m. It is interest-

Distributions fitted to the Poissonand the Greenwood-Yule curve.n = degrees of freedom; PN =

probability calculated according toNewbold.

ing that the diurnal variation is the same for

errors in figures read off directly as for errors in

figures that had required calculation.

Results of Main InvestigationThe material, covering the years 1912 to 1931,

includes about 175,000 entries. About 75,000errors were made in the entries. The greatestnumber of mistakes were made during the nightshift, the next greatest during the afternoon shift,and the lowest during the morning shift. The gaspressure columns are not included in these figures.The distribution of the errors during the 24 hours

is seen from the diagram in Fig. 1. The curve has

two distinct peaks, one at 3 a.m. and another at

3 p.m. A distinct depression is observed between

106

Sundays0

2345

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

DIURNAL VARIATION IN MENTAL PERFORMANCE

p00

Ar

4poo-1.1

%,,--.,. I

acoo

6 10 /4 /8 22

FIG. 1.-Distribution of the errors in the

1931 in 24 hours of the day and night. The

the abscissa and the errors along the ordinate.

5 p.m. and 7 p.m. and another one, not so distinct,

between 10 p.m. and p.m. Comparison

day curve from 6 a.m. to 6 p.m. with the

curve from 6 p.m. to 6 a.m. shows that the

a similar course, but that the fluctuation is

greater in the night curve.

It is remarkable that the curve is so little influenced

by the changes of shift at 6 a.m., 2 p.m., and 10

Because of the size of the series and the regular

course of the curve, it would be superfluous

make any calculations of the statistical significance.

In Table 4 the material for each year is analysed

separately. It is seen there that the two peaks

year after year. In 13 out of the 20 years the greatest

number of errors, the nocturnal peak, occurred

3 a.m., in six years at 2 a.m., and in one

4 a.m. The day peak was somewhat more fluctuating.

Various features of the material were analysed.

The distribution of the errors made in recording

the height of the barometer showed two peaks

3 a.m. and 3 p.m. This is worthy of note,

the atmospheric pressure is not influenced

changes in the production and consumption of

As already mentioned, by studying the

writing we were able to follow the different

who wrote in the ledgers. During the

question, three persons, A, C, and D, were

sible for most of the records. The errors made

these persons are further analysed in Fig. 2.

the material is divided into two parts, the material

from the years 1915 to 1919 and the material

the years 1920 to 1924. All three persons showed

distinct nocturnal peak in both the periods.

highest peaks occurred between and 4

six curves.

These two periods were chosen to see whether

change from the 56- to the 48-hour week

20 had any effect on the number

Fig. 2 shows that all three persons made

mistakes after their working hours were shortened.

The reduction was not due to the shorter

time perse. This is seen from Table 4 which

the total number of mistakes per year for each

Table 4 shows a steady reduction in the

oferrors per year. This reduction stopped

1920 to 1922, when the curve first ran a

course, and then showed a tendency to rise.

nation of the curve in Fig. 2 shows that

about the same course before and after the

of the working day. Subject D has a more

nocturnal peak before the reduction.

TABLE 4YEARLY DISTRIBUTION OF ERRORS ACCORDING TO TIME OF DAY AND NIGHT

1912 1913 1914 1915 1916 1917 1918 19191920 1921 1924 1925 1926 1927 1928 1929 1930 1931 Sum06 345 172 186 197 203 110 127 179 80 105 104 53 49 101 161 122 131 173 149 1172-86407 212 129 178 166 143 100 94 124 66 72 67 59 55 60 149 93 93 125 157 1152-25708 245 171 183 145 136 82 89 141 64 73 88 76 53 80 113 92 81 138 130 1372-31709 217 108 138 172 128 81 65 109 56 67 65 65 64 74 123 124 95 156 152 1332-19210 221 141 147 176 175 70 97 137 65 93 47 97 63 90 156 132 129 161 148 1542X49911 191 124 145 136 118 91 94 146 52 70 76 75 88 111 141 150 122 127 142 1292-32812 210 133 174 142 165 98 89 136 78 87 89 98 95 120 157 135 143 156 161 1632-62913 260 155 160 196 174 85 109 116 77 96 69 127 89 147 180, 165 135 172 163 1642-83914 332 180 178 217 184 112 125 123 103 139 105 184 95 203 2391 213 191 180 200 1803-48315 415 245 224 215 182 164 126 158 103 124 101 171 65 141 209 182 228 220 223 1903-68616 242 136 177 150 144 123 93 135 99 92 72 137 88 109 151 164 176 174 195 1382-79517 197 131 164 125 178 132 113 134 79 77 80 110 114 128 151 137 140 149 135 1592-63318 186 151 163 150 139 127 116 153 91 105 84 122 92 115 148 148 144 156 156 1322-67819 212 148 179 146 130 129 88 149 112 72 87 113 89 146 145 150 145 149 160 1362-68520 209 141 183 150 142 1011 112 148 91 91 84 110 80 130 167 150 173 150 175 140 2-72721 227 196 187 166 1801351 114 155 108 92 86 118 80 119 167 175 174 221 181 129 3*01022 313 175 194 204 148 132 143 171 84 149 115 143 96 132 214 181 172 202 173 1533-29423 273 202 221 178 162 142 143 198 95 113 96 110 87 126 138 190 148 177 187 1563-14224 267 171 217 199 207 167 155 192 101 122 114 140 105 125 184 177 185 207 211 1803-42601 382 190 279 254 193 143 181 252 143 132 128 156 130 138 193 202 221 247 256 1793-99902 433 214 268 293 248 196 238 247 188 187 158 221 166 218 260 224 262 250 277 2314-77903 582 246 359 308 280 182 309 354 183 159 202 299 131 199 313 276 292 252 289 2215-43604 394 222 315 261 285 167 295 268 142 133 192 213 103 156 209 197 200 198 235 156 4-34105 256 157 238 156 190 187 135 210 79 47 124 132 117 125 130 118 141 154 109 832-888::a 6,821 4,038 4,857 4,502 4,234 3,076 3,250 4,145 2,339 2,477 2,433 3,129 2,194 3,0934,1981 3,897 3,9211 4,294 4,364 3,675 74,927

107

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

BRITISH JOURNAL OF INDUSTRIAL MEDICINE

any systematic change in the fluctuation during thecourse of a night-shift week.The curves for the winter months November to

January and for the summer months May to Julywere compared for each of the three persons. As nodifference was noted, the figures are not consideredhere.The figures from another gasworks and from the

electrical control room of a paper mill are shown inTable 5. As seen there, both these series showedthe same peaks in errors at the same times of theday and night. The figures from the gasworks aretaken from the ledgers for 1944, 1945, and 1947.

100

10

FiG. 2.-Distribution of errors made by each of three employees(A, C, and D) at a gasworks.The bold curve is for the years 1915 to 1919 (56-hour week),

and the light curve for 1920 to 1924 (48-hour week).The hour is plotted along the abscissa and the number of

errors per shift along the ordinate.

As yet we have neglected the question of whetherthe propensity to error varies during the course ofa night-shift week. To examine this question, wetook the years 1915 to 1919 and compared the firstthree and the last three of the seven successive nightshifts done at that time. Fig. 3 shows that all threepersons evidenced remarkably little difference inperformance during the first and second halves ofthe week. After the shortening of the working weekin 1920, different types of shift were tried. Duringthe years 1922 to 1926 a free day was insertedbetween the first three and the last three nights of aweek of night work. It is interesting to see whetherthis had any influence on the high frequency oferrors during the night shift. As seen from Fig. 3,there was no essential difference in the course ofthe curve during the first and last nights of a night-shift week. The shortening of the working hoursdid not have any distinct influence on these curves.Nor did the figures for the separate persons show

200a

70o

VU

I22 24 2 4 6

Fio. 3.-Distribution throughout night-shift week of errors made bythree employees (A, C, and D) at a gasworks.The columns on the left cover the years 1915 to 1919, during

which period the men worked seven night shifts in succession.The columns on the right cover the years 1922 to 1926, duringwhich a free day was inserted between the first three and thelast three nights of a night-shift week.The bold curve shows the first three nights of night-shift

weeks, the fine curve the last three nights of night-shift weeks.The hour is plotted along the abscissa and the errors along the

ordinate.

CI05.

02.A A

D

a~~~~~~~~

4422 214 2? 4 -6

I n 9 0 a 0 I

108

I v v w

c

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

DIURNAL VARIATION IN MENTAL PERFORMANCE

TABLE 5

NUMBER OF ERRORS DISTRIBUTED ACCORDING TOTIME OF DAY AND NIGHT AT GASWORKS I, GASWORKS

II, AND PAPER MILL

Gasworks I

Hour Registration Sum of Gasworks II Paper Millof ErrorsBarometer fromHeight Table 4

06 23 2-864 147 8207 83 2-257 204 4008 84 2 317 229 3709 68 2-192 243 4110 83 2*499 231 5211 73 2-328 241 5712 99 2-629 249 6613 86 2-839 240 5614 127 3-483 297 8615 139 3-686 326 12016 92 2-795 283 10617 109 2-633 308 9818 105 2-678 271 9819 91 2*685 278 8020 98 2-727 290 8621 108 3-010 258 9522 114 3-294 390 10823 107 3-142 331 10724 105 3-426 393 10201 102 3-999 418 11602 121 4-779 413 10003 162 5-436 559 13304 113 4*341 441 11605 58 2-888 377 96

They cover about 26,300 entries containing 7,417errors. The figures for the paper mill are takenfrom the year 1945 and cover about 8,800 entriescontaining 2,078 errors.

DiscussionThe results reported here are based on a particular

performance repeated every hour of the day andnight for several years. The performance consistsof making note of the figures given on certaininstruments and recording some of them directly inledgers and others after calculation. The errorsmade in this test were undoubtedly caused by manythings. It may be assumed that variations in qualityof mental performance is one of them. As seenfrom the results, the frequency of error fluctuatesin a remarkably regular manner throughout the dayand night. Several sources of error, however, mustbe considered. The fluctuations may be due tochanges in the kind of work. They may be due tocyclic changes in the environment, for example, intemperature or degree of illumination. All thesesources of error must be eliminated before it ispossible to refer the variation in errors to variationswithin the person.The work dealt with in this study is of practically

the same nature throughout the 24 hours. It isconceivable, of course, that the test becomes moredifficult when there is a large consumption of gas,inasmuch as the figures fluctuate more, making thecalculation a little more difficult. But this would

not explain the curve here, for the gas consumptionis least during the night. Any doubt as to this isdispelled by the fact that the diurnal curve forerrors in noting the height of the barometer runsthe same course.The question arises whether the variation between

light and darkness during day and night might beinfluential. During the winter months Novemberto January it is dark during the whole night shift.During the summer months May to July it is lightfor the greater part of this shift. This did not causeany difference in the course of the nocturnal curve.Furthermore, it is darkest at midnight but theerrors showed a distinct peak at 3 a.m.

Since no fluctuations in the environment can befound to explain the diurnal fluctuation in errors, itmust be assumed that the latter is due to variationin the individual in question, to variation in hiscapacity for mental performance. It is reasonableto assume that the curve reflects the degree of wake-fulness during the day and night. The fact that thesame type of variation is shown year after year bydifferent persons and is seen in different seriesshows that the degree of wakefulness fluctuates ina highly consistent fashion during the 24 hours ofthe day and night. The large increase of errorsduring the night corresponds well with the feelingwe all have of being more tired at night than duringthe day, something which is particularly noticeableduring nocturnal vigils. Persons doing the type ofnight work described in this study often say thatthey are sometimes so tired between 2 and 3 a.m.that they fall asleep when they are supposed towrite. Menzel (1950) made systematic inquiries of210 shift workers with the same times for changingshift as our subjects. The answers to the inquiryas to the most difficult time of the night to workdistributed themselves in a Gaussian curve with apeak at 3 a.m. This agrees with our results.

Because of the large number of observations, thedegree of fatigue at each point being " tested "

about 7,000 times, the results are convincing, atleast as regards the subjects in question. This isespecially true in view of the fact that a series fromanother gasworks and a paper mill gave the sameresult. It must be remembered, however, that thesubjects were few in number, the bulk of thematerial in the main study being furnished by threepersons. Furthermore, these persons had a mono-tonous scheme of work, with a change of shift at6 a.m., 2 p.m., and 10 p.m. This is also true of theother two series. It is possible and even probablethat persons with other habits of living and otherhours of work would show another diurnal curve.It would be very interesting to compare our results

109

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from

BRITISH JOURNAL OF INDUSTRIAL MEDICINE

with those of studies of groups with other workinghours. We have searched a great deal but have notbeen able to find suitable material for comparison.As mentioned earlier, a series of biological

processes has been found to vary according to adiurnal rhythm. As the situation is now, it is idleto speculate on which of these rhythms are primaryand which are secondary.The practical significance of this fluctuation in

mental performance is obvious as regards industrialworkers, transport employees, and the like. Itseems that one cannot expect a person to have thesame capacity during the night as during the day.It is interesting to note that the shortening of dailyhours of work in 1920 did not have a great influenceon the diurnal variation. It is likewise interestingthat the fluctuation seems to follow the same curveduring the last three nights of a night shift as duringthe first three. It seems as though seven days werenot enough for the organism to adapt itself tonight work. Holmgren and Swensson showed thatanimals needed about a week to adjust themselvesto a change in diurnal schedule. This in no waycontradicts our observation, however, for their

experimental animals and our human subjects werefaced with quite different situations. The animalsonly had to adjust their activity to the new order intheir society. The man changing to a night shift,however, has to change his personal routine to onewhich is quite different from his fellow men, whogo on living in the same old way.

REFERENCESBjerner, B. (1949). Acta physiol. scand., Suppl. 65.

, and Swensson, A. (1953). Acta med. scand., Suppl. 278, p. 102.-, Holm, A., and Swensson, A. (1948). Statens offentliga

Utredningar. Nr 51. Stockholm.Browne, R. C. (1949). Occup. Psychol., 23, 121.Freeman, G. L., and Hovland, C. I. (1934). Psychol. Bull., 31, 777.Greenwood, M., and Yule, G. U. (1920). J. roy. Stat. Soc., 83,

part II, 255.Holmgren, H. (1936). Studien tiber 24-Stundenrhythmische Varia-

tionen des Darm-, Lungen- und Leberfetts. Acta med.scand., Suppl. 74.and Swensson, 'A. (1953). Ibid., Suppl. 278, p. 71.

Industrial Fatigue Research Board Reports (1919-22).Jores, A. (1937). Tab. biol., Amst., 14, 77.Kleitman, N. (1939). Sleep and Wakefulness. University of Chicago

Press.-(1949). Physiol. Rev., 29, 1.Menzel, W. (1950). Arbeitsphysiologie, 14, 304.Ministry of Munitions (1918). Health of Munition Workers Com-

mittee. Final Report (Cd. 9065). H.M.S.O., London.Newbold, E. M. (1926). Industrial Fatigue Research Board, Rep. 34.Vernon, H. M. (1921). Industrial Fatigue and Efficiency. Routledge,

London-(1940). The Health and Efficiency ofMunition Workers. Oxford

University Press, London.

110

on Novem

ber 20, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.12.2.103 on 1 A

pril 1955. Dow

nloaded from